fs/ioctl.c

   1 /*
   2  *  linux/fs/ioctl.c
   3  *
   4  *  Copyright (C) 1991, 1992  Linus Torvalds
   5  */
   6
   7 #include <linux/syscalls.h>
   8 #include <linux/mm.h>
   9 #include <linux/smp_lock.h>
  10 #include <linux/capability.h>
  11 #include <linux/file.h>
  12 #include <linux/fs.h>
  13 #include <linux/security.h>
  14 #include <linux/module.h>
  15 #include <linux/uaccess.h>
  16 #include <linux/writeback.h>
  17 #include <linux/buffer_head.h>
  18 #include <linux/falloc.h>
  19
  20 #include <asm/ioctls.h>
  21
  22 /* So that the fiemap access checks can't overflow on 32 bit machines. */
  23 #define FIEMAP_MAX_EXTENTS      (UINT_MAX / sizeof(struct fiemap_extent))
  24
  25 /**
  26  * vfs_ioctl - call filesystem specific ioctl methods
  27  * @filp:       open file to invoke ioctl method on
  28  * @cmd:        ioctl command to execute
  29  * @arg:        command-specific argument for ioctl
  30  *
  31  * Invokes filesystem specific ->unlocked_ioctl, if one exists; otherwise
  32  * invokes filesystem specific ->ioctl method.  If neither method exists,
  33  * returns -ENOTTY.
  34  *
  35  * Returns 0 on success, -errno on error.
  36  */
  37 static long vfs_ioctl(struct file *filp, unsigned int cmd,
  38                       unsigned long arg)
  39 {
  40         int error = -ENOTTY;
  41
  42         if (!filp->f_op)
  43                 goto out;
  44
  45         if (filp->f_op->unlocked_ioctl) {
  46                 error = filp->f_op->unlocked_ioctl(filp, cmd, arg);
  47                 if (error == -ENOIOCTLCMD)
  48                         error = -EINVAL;
  49                 goto out;
  50         } else if (filp->f_op->ioctl) {
  51                 lock_kernel();
  52                 error = filp->f_op->ioctl(filp->f_path.dentry->d_inode,
  53                                           filp, cmd, arg);
  54                 unlock_kernel();
  55         }
  56
  57  out:
  58         return error;
  59 }
  60
  61 static int ioctl_fibmap(struct file *filp, int __user *p)
  62 {
  63         struct address_space *mapping = filp->f_mapping;
  64         int res, block;
  65
  66         /* do we support this mess? */
  67         if (!mapping->a_ops->bmap)
  68                 return -EINVAL;
  69         if (!capable(CAP_SYS_RAWIO))
  70                 return -EPERM;
  71         res = get_user(block, p);
  72         if (res)
  73                 return res;
  74         res = mapping->a_ops->bmap(mapping, block);
  75         return put_user(res, p);
  76 }
  77
  78 /**
  79  * fiemap_fill_next_extent - Fiemap helper function
  80  * @fieinfo:    Fiemap context passed into ->fiemap
  81  * @logical:    Extent logical start offset, in bytes
  82  * @phys:       Extent physical start offset, in bytes
  83  * @len:        Extent length, in bytes
  84  * @flags:      FIEMAP_EXTENT flags that describe this extent
  85  *
  86  * Called from file system ->fiemap callback. Will populate extent
  87  * info as passed in via arguments and copy to user memory. On
  88  * success, extent count on fieinfo is incremented.
  89  *
  90  * Returns 0 on success, -errno on error, 1 if this was the last
  91  * extent that will fit in user array.
  92  */
  93 #define SET_UNKNOWN_FLAGS       (FIEMAP_EXTENT_DELALLOC)
  94 #define SET_NO_UNMOUNTED_IO_FLAGS       (FIEMAP_EXTENT_DATA_ENCRYPTED)
  95 #define SET_NOT_ALIGNED_FLAGS   (FIEMAP_EXTENT_DATA_TAIL|FIEMAP_EXTENT_DATA_INLINE)
  96 int fiemap_fill_next_extent(struct fiemap_extent_info *fieinfo, u64 logical,
  97                             u64 phys, u64 len, u32 flags)
  98 {
  99         struct fiemap_extent extent;
 100         struct fiemap_extent *dest = fieinfo->fi_extents_start;
 101
 102         /* only count the extents */
 103         if (fieinfo->fi_extents_max == 0) {
 104                 fieinfo->fi_extents_mapped++;
 105                 return (flags & FIEMAP_EXTENT_LAST) ? 1 : 0;
 106         }
 107
 108         if (fieinfo->fi_extents_mapped >= fieinfo->fi_extents_max)
 109                 return 1;
 110
 111         if (flags & SET_UNKNOWN_FLAGS)
 112                 flags |= FIEMAP_EXTENT_UNKNOWN;
 113         if (flags & SET_NO_UNMOUNTED_IO_FLAGS)
 114                 flags |= FIEMAP_EXTENT_ENCODED;
 115         if (flags & SET_NOT_ALIGNED_FLAGS)
 116                 flags |= FIEMAP_EXTENT_NOT_ALIGNED;
 117
 118         memset(&extent, 0, sizeof(extent));
 119         extent.fe_logical = logical;
 120         extent.fe_physical = phys;
 121         extent.fe_length = len;
 122         extent.fe_flags = flags;
 123
 124         dest += fieinfo->fi_extents_mapped;
 125         if (copy_to_user(dest, &extent, sizeof(extent)))
 126                 return -EFAULT;
 127
 128         fieinfo->fi_extents_mapped++;
 129         if (fieinfo->fi_extents_mapped == fieinfo->fi_extents_max)
 130                 return 1;
 131         return (flags & FIEMAP_EXTENT_LAST) ? 1 : 0;
 132 }
 133 EXPORT_SYMBOL(fiemap_fill_next_extent);
 134
 135 /**
 136  * fiemap_check_flags - check validity of requested flags for fiemap
 137  * @fieinfo:    Fiemap context passed into ->fiemap
 138  * @fs_flags:   Set of fiemap flags that the file system understands
 139  *
 140  * Called from file system ->fiemap callback. This will compute the
 141  * intersection of valid fiemap flags and those that the fs supports. That
 142  * value is then compared against the user supplied flags. In case of bad user
 143  * flags, the invalid values will be written into the fieinfo structure, and
 144  * -EBADR is returned, which tells ioctl_fiemap() to return those values to
 145  * userspace. For this reason, a return code of -EBADR should be preserved.
 146  *
 147  * Returns 0 on success, -EBADR on bad flags.
 148  */
 149 int fiemap_check_flags(struct fiemap_extent_info *fieinfo, u32 fs_flags)
 150 {
 151         u32 incompat_flags;
 152
 153         incompat_flags = fieinfo->fi_flags & ~(FIEMAP_FLAGS_COMPAT & fs_flags);
 154         if (incompat_flags) {
 155                 fieinfo->fi_flags = incompat_flags;
 156                 return -EBADR;
 157         }
 158         return 0;
 159 }
 160 EXPORT_SYMBOL(fiemap_check_flags);
 161
 162 static int fiemap_check_ranges(struct super_block *sb,
 163                                u64 start, u64 len, u64 *new_len)
 164 {
 165         *new_len = len;
 166
 167         if (len == 0)
 168                 return -EINVAL;
 169
 170         if (start > sb->s_maxbytes)
 171                 return -EFBIG;
 172
 173         /*
 174          * Shrink request scope to what the fs can actually handle.
 175          */
 176         if ((len > sb->s_maxbytes) ||
 177             (sb->s_maxbytes - len) < start)
 178                 *new_len = sb->s_maxbytes - start;
 179
 180         return 0;
 181 }
 182
 183 static int ioctl_fiemap(struct file *filp, unsigned long arg)
 184 {
 185         struct fiemap fiemap;
 186         struct fiemap_extent_info fieinfo = { 0, };
 187         struct inode *inode = filp->f_path.dentry->d_inode;
 188         struct super_block *sb = inode->i_sb;
 189         u64 len;
 190         int error;
 191
 192         if (!inode->i_op->fiemap)
 193                 return -EOPNOTSUPP;
 194
 195         if (copy_from_user(&fiemap, (struct fiemap __user *)arg,
 196                            sizeof(struct fiemap)))
 197                 return -EFAULT;
 198
 199         if (fiemap.fm_extent_count > FIEMAP_MAX_EXTENTS)
 200                 return -EINVAL;
 201
 202         error = fiemap_check_ranges(sb, fiemap.fm_start, fiemap.fm_length,
 203                                     &len);
 204         if (error)
 205                 return error;
 206
 207         fieinfo.fi_flags = fiemap.fm_flags;
 208         fieinfo.fi_extents_max = fiemap.fm_extent_count;
 209         fieinfo.fi_extents_start = (struct fiemap_extent *)(arg + sizeof(fiemap));
 210
 211         if (fiemap.fm_extent_count != 0 &&
 212             !access_ok(VERIFY_WRITE, fieinfo.fi_extents_start,
 213                        fieinfo.fi_extents_max * sizeof(struct fiemap_extent)))
 214                 return -EFAULT;
 215
 216         if (fieinfo.fi_flags & FIEMAP_FLAG_SYNC)
 217                 filemap_write_and_wait(inode->i_mapping);
 218
 219         error = inode->i_op->fiemap(inode, &fieinfo, fiemap.fm_start, len);
 220         fiemap.fm_flags = fieinfo.fi_flags;
 221         fiemap.fm_mapped_extents = fieinfo.fi_extents_mapped;
 222         if (copy_to_user((char *)arg, &fiemap, sizeof(fiemap)))
 223                 error = -EFAULT;
 224
 225         return error;
 226 }
 227
 228 #ifdef CONFIG_BLOCK
 229
 230 #define blk_to_logical(inode, blk) (blk << (inode)->i_blkbits)
 231 #define logical_to_blk(inode, offset) (offset >> (inode)->i_blkbits);
 232
 233 /**
 234  * __generic_block_fiemap - FIEMAP for block based inodes (no locking)
 235  * @inode - the inode to map
 236  * @arg - the pointer to userspace where we copy everything to
 237  * @get_block - the fs's get_block function
 238  *
 239  * This does FIEMAP for block based inodes.  Basically it will just loop
 240  * through get_block until we hit the number of extents we want to map, or we
 241  * go past the end of the file and hit a hole.
 242  *
 243  * If it is possible to have data blocks beyond a hole past @inode->i_size, then
 244  * please do not use this function, it will stop at the first unmapped block
 245  * beyond i_size.
 246  *
 247  * If you use this function directly, you need to do your own locking. Use
 248  * generic_block_fiemap if you want the locking done for you.
 249  */
 250
 251 int __generic_block_fiemap(struct inode *inode,
 252                            struct fiemap_extent_info *fieinfo, u64 start,
 253                            u64 len, get_block_t *get_block)
 254 {
 255         struct buffer_head tmp;
 256         unsigned int start_blk;
 257         long long length = 0, map_len = 0;
 258         u64 logical = 0, phys = 0, size = 0;
 259         u32 flags = FIEMAP_EXTENT_MERGED;
 260         int ret = 0, past_eof = 0, whole_file = 0;
 261
 262         if ((ret = fiemap_check_flags(fieinfo, FIEMAP_FLAG_SYNC)))
 263                 return ret;
 264
 265         start_blk = logical_to_blk(inode, start);
 266
 267         length = (long long)min_t(u64, len, i_size_read(inode));
 268         if (length < len)
 269                 whole_file = 1;
 270
 271         map_len = length;
 272
 273         do {
 274                 /*
 275                  * we set b_size to the total size we want so it will map as
 276                  * many contiguous blocks as possible at once
 277                  */
 278                 memset(&tmp, 0, sizeof(struct buffer_head));
 279                 tmp.b_size = map_len;
 280
 281                 ret = get_block(inode, start_blk, &tmp, 0);
 282                 if (ret)
 283                         break;
 284
 285                 /* HOLE */
 286                 if (!buffer_mapped(&tmp)) {
 287                         length -= blk_to_logical(inode, 1);
 288                         start_blk++;
 289
 290                         /*
 291                          * we want to handle the case where there is an
 292                          * allocated block at the front of the file, and then
 293                          * nothing but holes up to the end of the file properly,
 294                          * to make sure that extent at the front gets properly
 295                          * marked with FIEMAP_EXTENT_LAST
 296                          */
 297                         if (!past_eof &&
 298                             blk_to_logical(inode, start_blk) >=
 299                             blk_to_logical(inode, 0)+i_size_read(inode))
 300                                 past_eof = 1;
 301
 302                         /*
 303                          * first hole after going past the EOF, this is our
 304                          * last extent
 305                          */
 306                         if (past_eof && size) {
 307                                 flags = FIEMAP_EXTENT_MERGED|FIEMAP_EXTENT_LAST;
 308                                 ret = fiemap_fill_next_extent(fieinfo, logical,
 309                                                               phys, size,
 310                                                               flags);
 311                                 break;
 312                         }
 313
 314                         /* if we have holes up to/past EOF then we're done */
 315                         if (length <= 0 || past_eof)
 316                                 break;
 317                 } else {
 318                         /*
 319                          * we have gone over the length of what we wanted to
 320                          * map, and it wasn't the entire file, so add the extent
 321                          * we got last time and exit.
 322                          *
 323                          * This is for the case where say we want to map all the
 324                          * way up to the second to the last block in a file, but
 325                          * the last block is a hole, making the second to last
 326                          * block FIEMAP_EXTENT_LAST.  In this case we want to
 327                          * see if there is a hole after the second to last block
 328                          * so we can mark it properly.  If we found data after
 329                          * we exceeded the length we were requesting, then we
 330                          * are good to go, just add the extent to the fieinfo
 331                          * and break
 332                          */
 333                         if (length <= 0 && !whole_file) {
 334                                 ret = fiemap_fill_next_extent(fieinfo, logical,
 335                                                               phys, size,
 336                                                               flags);
 337                                 break;
 338                         }
 339
 340                         /*
 341                          * if size != 0 then we know we already have an extent
 342                          * to add, so add it.
 343                          */
 344                         if (size) {
 345                                 ret = fiemap_fill_next_extent(fieinfo, logical,
 346                                                               phys, size,
 347                                                               flags);
 348                                 if (ret)
 349                                         break;
 350                         }
 351
 352                         logical = blk_to_logical(inode, start_blk);
 353                         phys = blk_to_logical(inode, tmp.b_blocknr);
 354                         size = tmp.b_size;
 355                         flags = FIEMAP_EXTENT_MERGED;
 356
 357                         length -= tmp.b_size;
 358                         start_blk += logical_to_blk(inode, size);
 359
 360                         /*
 361                          * If we are past the EOF, then we need to make sure as
 362                          * soon as we find a hole that the last extent we found
 363                          * is marked with FIEMAP_EXTENT_LAST
 364                          */
 365                         if (!past_eof &&
 366                             logical+size >=
 367                             blk_to_logical(inode, 0)+i_size_read(inode))
 368                                 past_eof = 1;
 369                 }
 370                 cond_resched();
 371         } while (1);
 372
 373         /* if ret is 1 then we just hit the end of the extent array */
 374         if (ret == 1)
 375                 ret = 0;
 376
 377         return ret;
 378 }
 379 EXPORT_SYMBOL(__generic_block_fiemap);
 380
 381 /**
 382  * generic_block_fiemap - FIEMAP for block based inodes
 383  * @inode: The inode to map
 384  * @fieinfo: The mapping information
 385  * @start: The initial block to map
 386  * @len: The length of the extect to attempt to map
 387  * @get_block: The block mapping function for the fs
 388  *
 389  * Calls __generic_block_fiemap to map the inode, after taking
 390  * the inode's mutex lock.
 391  */
 392
 393 int generic_block_fiemap(struct inode *inode,
 394                          struct fiemap_extent_info *fieinfo, u64 start,
 395                          u64 len, get_block_t *get_block)
 396 {
 397         int ret;
 398         mutex_lock(&inode->i_mutex);
 399         ret = __generic_block_fiemap(inode, fieinfo, start, len, get_block);
 400         mutex_unlock(&inode->i_mutex);
 401         return ret;
 402 }
 403 EXPORT_SYMBOL(generic_block_fiemap);
 404
 405 #endif  /*  CONFIG_BLOCK  */
 406
 407 /*
 408  * This provides compatibility with legacy XFS pre-allocation ioctls
 409  * which predate the fallocate syscall.
 410  *
 411  * Only the l_start, l_len and l_whence fields of the 'struct space_resv'
 412  * are used here, rest are ignored.
 413  */
 414 int ioctl_preallocate(struct file *filp, void __user *argp)
 415 {
 416         struct inode *inode = filp->f_path.dentry->d_inode;
 417         struct space_resv sr;
 418
 419         if (copy_from_user(&sr, argp, sizeof(sr)))
 420                 return -EFAULT;
 421
 422         switch (sr.l_whence) {
 423         case SEEK_SET:
 424                 break;
 425         case SEEK_CUR:
 426                 sr.l_start += filp->f_pos;
 427                 break;
 428         case SEEK_END:
 429                 sr.l_start += i_size_read(inode);
 430                 break;
 431         default:
 432                 return -EINVAL;
 433         }
 434
 435         return do_fallocate(filp, FALLOC_FL_KEEP_SIZE, sr.l_start, sr.l_len);
 436 }
 437
 438 static int file_ioctl(struct file *filp, unsigned int cmd,
 439                 unsigned long arg)
 440 {
 441         struct inode *inode = filp->f_path.dentry->d_inode;
 442         int __user *p = (int __user *)arg;
 443
 444         switch (cmd) {
 445         case FIBMAP:
 446                 return ioctl_fibmap(filp, p);
 447         case FIONREAD:
 448                 return put_user(i_size_read(inode) - filp->f_pos, p);
 449         case FS_IOC_RESVSP:
 450         case FS_IOC_RESVSP64:
 451                 return ioctl_preallocate(filp, p);
 452         }
 453
 454         return vfs_ioctl(filp, cmd, arg);
 455 }
 456
 457 static int ioctl_fionbio(struct file *filp, int __user *argp)
 458 {
 459         unsigned int flag;
 460         int on, error;
 461
 462         error = get_user(on, argp);
 463         if (error)
 464                 return error;
 465         flag = O_NONBLOCK;
 466 #ifdef __sparc__
 467         /* SunOS compatibility item. */
 468         if (O_NONBLOCK != O_NDELAY)
 469                 flag |= O_NDELAY;
 470 #endif
 471         spin_lock(&filp->f_lock);
 472         if (on)
 473                 filp->f_flags |= flag;
 474         else
 475                 filp->f_flags &= ~flag;
 476         spin_unlock(&filp->f_lock);
 477         return error;
 478 }
 479
 480 static int ioctl_fioasync(unsigned int fd, struct file *filp,
 481                           int __user *argp)
 482 {
 483         unsigned int flag;
 484         int on, error;
 485
 486         error = get_user(on, argp);
 487         if (error)
 488                 return error;
 489         flag = on ? FASYNC : 0;
 490
 491         /* Did FASYNC state change ? */
 492         if ((flag ^ filp->f_flags) & FASYNC) {
 493                 if (filp->f_op && filp->f_op->fasync)
 494                         /* fasync() adjusts filp->f_flags */
 495                         error = filp->f_op->fasync(fd, filp, on);
 496                 else
 497                         error = -ENOTTY;
 498         }
 499         return error < 0 ? error : 0;
 500 }
 501
 502 static int ioctl_fsfreeze(struct file *filp)
 503 {
 504         struct super_block *sb = filp->f_path.dentry->d_inode->i_sb;
 505
 506         if (!capable(CAP_SYS_ADMIN))
 507                 return -EPERM;
 508
 509         /* If filesystem doesn't support freeze feature, return. */
 510         if (sb->s_op->freeze_fs == NULL)
 511                 return -EOPNOTSUPP;
 512
 513         /* If a blockdevice-backed filesystem isn't specified, return. */
 514         if (sb->s_bdev == NULL)
 515                 return -EINVAL;
 516
 517         /* Freeze */
 518         sb = freeze_bdev(sb->s_bdev);
 519         if (IS_ERR(sb))
 520                 return PTR_ERR(sb);
 521         return 0;
 522 }
 523
 524 static int ioctl_fsthaw(struct file *filp)
 525 {
 526         struct super_block *sb = filp->f_path.dentry->d_inode->i_sb;
 527
 528         if (!capable(CAP_SYS_ADMIN))
 529                 return -EPERM;
 530
 531         /* If a blockdevice-backed filesystem isn't specified, return EINVAL. */
 532         if (sb->s_bdev == NULL)
 533                 return -EINVAL;
 534
 535         /* Thaw */
 536         return thaw_bdev(sb->s_bdev, sb);
 537 }
 538
 539 /*
 540  * When you add any new common ioctls to the switches above and below
 541  * please update compat_sys_ioctl() too.
 542  *
 543  * do_vfs_ioctl() is not for drivers and not intended to be EXPORT_SYMBOL()'d.
 544  * It's just a simple helper for sys_ioctl and compat_sys_ioctl.
 545  */
 546 int do_vfs_ioctl(struct file *filp, unsigned int fd, unsigned int cmd,
 547              unsigned long arg)
 548 {
 549         int error = 0;
 550         int __user *argp = (int __user *)arg;
 551
 552         switch (cmd) {
 553         case FIOCLEX:
 554                 set_close_on_exec(fd, 1);
 555                 break;
 556
 557         case FIONCLEX:
 558                 set_close_on_exec(fd, 0);
 559                 break;
 560
 561         case FIONBIO:
 562                 error = ioctl_fionbio(filp, argp);
 563                 break;
 564
 565         case FIOASYNC:
 566                 error = ioctl_fioasync(fd, filp, argp);
 567                 break;
 568
 569         case FIOQSIZE:
 570                 if (S_ISDIR(filp->f_path.dentry->d_inode->i_mode) ||
 571                     S_ISREG(filp->f_path.dentry->d_inode->i_mode) ||
 572                     S_ISLNK(filp->f_path.dentry->d_inode->i_mode)) {
 573                         loff_t res =
 574                                 inode_get_bytes(filp->f_path.dentry->d_inode);
 575                         error = copy_to_user((loff_t __user *)arg, &res,
 576                                              sizeof(res)) ? -EFAULT : 0;
 577                 } else
 578                         error = -ENOTTY;
 579                 break;
 580
 581         case FIFREEZE:
 582                 error = ioctl_fsfreeze(filp);
 583                 break;
 584
 585         case FITHAW:
 586                 error = ioctl_fsthaw(filp);
 587                 break;
 588
 589         case FS_IOC_FIEMAP:
 590                 return ioctl_fiemap(filp, arg);
 591
 592         case FIGETBSZ:
 593         {
 594                 struct inode *inode = filp->f_path.dentry->d_inode;
 595                 int __user *p = (int __user *)arg;
 596                 return put_user(inode->i_sb->s_blocksize, p);
 597         }
 598
 599         default:
 600                 if (S_ISREG(filp->f_path.dentry->d_inode->i_mode))
 601                         error = file_ioctl(filp, cmd, arg);
 602                 else
 603                         error = vfs_ioctl(filp, cmd, arg);
 604                 break;
 605         }
 606         return error;
 607 }
 608
 609 SYSCALL_DEFINE3(ioctl, unsigned int, fd, unsigned int, cmd, unsigned long, arg)
 610 {
 611         struct file *filp;
 612         int error = -EBADF;
 613         int fput_needed;
 614
 615         filp = fget_light(fd, &fput_needed);
 616         if (!filp)
 617                 goto out;
 618
 619         error = security_file_ioctl(filp, cmd, arg);
 620         if (error)
 621                 goto out_fput;
 622
 623         error = do_vfs_ioctl(filp, fd, cmd, arg);
 624  out_fput:
 625         fput_light(filp, fput_needed);
 626  out:
 627         return error;
 628 }